Local AI
Local AI for document search
Document search at home is two pieces: a small search model that finds the right pages, and a chat model that writes the answer. We pair BAAI bge-m3 (about 1.2 GB extra) with the same chat models used for writing. AnythingLLM and Open WebUI are the friendly apps. Their official guides cover setup.
Document search
Runs wellA step below DeepSeek (July 2026)
The speed pick. Nearly as liked as the best local writer, and much quicker, because only a slice of it wakes up per word.
Mac mini M4 32 GB
Around $1,099 · estimate
Source: Apple M4 Mac mini memory options, 14 Oct 2026
Install guides from the makers
- Ollama's download page · Ollama · Mac, Windows, Linux
- LM Studio's app guide · LM Studio · Mac, Windows, Linux
- AnythingLLM's docs · AnythingLLM · Mac, Windows, Linux
#ad As an Amazon Associate, LocalAIBuddy earns from qualifying purchases. GMKtec, Minisforum, RunPod, and Vast.ai links use placeholder referral codes and may pay us. Prices checked 14 Oct 2026. Apple links on this page are not paid.
Questions people actually ask
Is this the same as uploading files to ChatGPT?
The idea is the same. The difference is that the files and the answers stay on your computer. Answer quality follows the chat model that fits.
How much extra memory do the files need?
The search model is about 1.2 GB. Your documents sit on disk. Leave a few extra gigabytes so the chat model still has room to answer.
Distinct answers
Search engines see only the computers where this job's answer is genuinely different. The other computers stay on the site and point here.